In [10]:
import scipy.stats as stats
from scipy.interpolate import CubicSpline
from scipy.interpolate import UnivariateSpline
from statsmodels.stats.power import TTestPower
from matplotlib.patches import Patch
%config InlineBackend.figure_format = 'retina'
plt.rc('figure', dpi=300)
analysis = TTestPower()
fig = plt.figure(figsize=(30, 12),constrained_layout=True)
gs = fig.add_gridspec(3, 5)
# fig.suptitle('Comparison of classification performance', fontsize=36, y=1.06)
sample_colors = np.array(sns.color_palette("rainbow", len(sample_sizes)))
logp_threshold = -np.log10(0.05)
for ci, task in enumerate(tasks):
for ri, method in enumerate(['spatial', 'topological', 'difference']):
ax = fig.add_subplot(gs[ri, ci])
nauc = []
for si, sample_size in enumerate(sample_sizes):
if method == 'difference':
tpr = TPR[task][f'N={sample_size}']['topological'].mean(1)
fpr = FPR[task][f'N={sample_size}']['topological'].mean(1)
bmt = BM[task][f'N={sample_size}']['topological']
bms = BM[task][f'N={sample_size}']['spatial']
bmdiff = bmt - bms
xlim = (min(ground_truth_effect_thresholds),max(ground_truth_effect_thresholds))
sns.lineplot(
x=ground_truth_effect_thresholds,
y=bmdiff.mean(1),
style=True,
color=np.append(sample_colors[si]/1.2, 1),
legend=False,
linewidth=2,
)
ax.fill_between(
ground_truth_effect_thresholds,
bmdiff.mean(1) - (stats.sem(bmdiff, axis=1)*1.96),
bmdiff.mean(1) + (stats.sem(bmdiff, axis=1)*1.96),
color = np.append(sample_colors[si], 0.3),
)
dtof = UnivariateSpline(ground_truth_effect_thresholds, fpr, s=0)
dtot = UnivariateSpline(ground_truth_effect_thresholds, tpr, s=0)
dtobmd = UnivariateSpline(ground_truth_effect_thresholds, bmdiff.mean(1), s=0)
nauc.append(
compute_normalized_partial_area_under_curve(
ground_truth_effect_thresholds,
bmdiff.mean(1),
get_cohen_d_from_bonferroni_corrected_p_value(0.05, 1000, 1),
get_cohen_d_from_bonferroni_corrected_p_value(0.05, 1000, 91282),
)
)
else:
tpr = TPR[task][f'N={sample_size}'][method].mean(1)
fpr = FPR[task][f'N={sample_size}'][method].mean(1)
bm = BM[task][f'N={sample_size}'][method]
sns.lineplot(
x=ground_truth_effect_thresholds,
y=bm.mean(1),
style=True,
color=np.append(sample_colors[si]/1.2, 1),
legend=False,
linewidth=2,
)
ax.fill_between(
ground_truth_effect_thresholds,
bm.mean(1) - (stats.sem(bm, axis=1)*1.96),
bm.mean(1) + (stats.sem(bm, axis=1)*1.96),
color = np.append(sample_colors[si], 0.3),
)
xlim = (min(ground_truth_effect_thresholds),max(ground_truth_effect_thresholds))
dtof = UnivariateSpline(ground_truth_effect_thresholds, fpr, s=0)
dtot = UnivariateSpline(ground_truth_effect_thresholds, tpr, s=0)
nauc.append(
compute_normalized_partial_area_under_curve(
ground_truth_effect_thresholds,
bm.mean(1),
get_cohen_d_from_bonferroni_corrected_p_value(0.05, 1000, 1),
get_cohen_d_from_bonferroni_corrected_p_value(0.05, 1000, 91282),
)
)
ax.set_ylim(-0.05,1.05)
leg = ax.legend(
handles=[Patch(facecolor=np.append(sample_colors[si], 1),) for si, sample_size in enumerate(sample_sizes)],
labels=['${:.1f}\%$'.format(100*x) for x in nauc],
loc='upper left',
ncol=1,
fontsize=18,
title="$AUC$",
title_fontsize=24,
labelspacing=1.,
handlelength=1.,
)
leg.set_zorder(30)
for patch in leg.get_patches():
patch.set_width(20)
patch.set_height(20)
patch.set_y(-4)
xlabel = ''
if ri == 2:
xlabel = 'Effect size threshold'
ax.set_xlabel(xlabel, fontsize=40)
ax.set_xlim(0.0,0.21)
ylabel = ''
if ci == 0:
ylabel = '{}\nTrue Positive Rate'.format(task)
axylim = ax.get_ylim()
ax.vlines(
x=float(get_cohen_d_from_bonferroni_corrected_p_value(0.05, 1000, 91282)),
ymin=axylim[0],
ymax=axylim[1],
linestyles='dashed',
colors=np.array([212,24,0])/255,
zorder = 21,
linewidth=2,
)
ax.fill_between(
[get_cohen_d_from_bonferroni_corrected_p_value(0.05, 1000, 91282), 0.21],
[axylim[0], axylim[0]],
[axylim[1], axylim[1]],
color = [0.7,0.7,0.7,0.6],
zorder = 20,
)
ax.vlines(
x=float(get_cohen_d_from_bonferroni_corrected_p_value(0.05, 1000, 1)),
ymin=axylim[0],
ymax=axylim[1],
linestyles='dashed',
colors=np.array([212,24,0])/255,
zorder = 21,
linewidth=2,
)
ax.fill_between(
[0, get_cohen_d_from_bonferroni_corrected_p_value(0.05, 1000, 1),],
[axylim[0], axylim[0]],
[axylim[1], axylim[1]],
color = [0.7,0.7,0.7,0.6],
zorder = 20,
)
ax.set_ylim(axylim)
ax.set_facecolor(np.array([234,234,242])/255)
ax.grid(color=(0.99,0.99,0.99,), linewidth=3)
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['bottom'].set_visible(False)
ax.spines['left'].set_visible(False)
ax.tick_params(axis='both', colors=(0.5,0.5,0.5), labelcolor=(0,0,0), direction='out')
fig.legend(
handles=[Patch(facecolor=np.append(sample_colors[si], 1),) for si, sample_size in enumerate(sample_sizes)],
labels=[f'N={sample_size}' for sample_size in sample_sizes],
loc='lower center',
bbox_to_anchor=(0.5, 1.0),
ncol=len(sample_sizes), fontsize=40,
)
plt.show()
